摘要
针对模糊C均值聚类(FCM)算法在样本集不理想的情况下会导致不理想的聚类结果,提出了一种改进方法。将某样本对于各个聚类的隶属度之和为1这一约束条件,改变为所有样本对各类的隶属度总和设为样本总数。实验表明,该方法用于人脑磁共振图像的分割时,分割准确度大大提高。
The fuzzy C- means clustering （FCM） algorithm gives the unperfect clustering results due to unperfect data samples. This paper proposes a modified FCM algorithm,in which the old constrained condition that the sum of membership values of a sample belonging to all clustering centers is 1,changes to that the sum of membership values of all sample belonging to all clustering centers is the number of all sample. The algorithm presented is applied to the problem of segmenting brain magnetic resonance image ,and tests show the accuracy of segmentation is improved.
出处
《电脑开发与应用》
2006年第4期10-11,共2页
Computer Development & Applications
基金
甘肃省自然科学基金(32S042B25014)资助
